Autocorrelation Analysis Pipeline

A MATLAB pipeline for analyzing neural calcium imaging data through autocorrelation analysis, oscillation detection, and statistical validation using block-bootstrapping.

Overview

This pipeline processes calcium imaging traces to identify oscillatory patterns in neural activity. It performs detrending, autocorrelation analysis, frequency domain analysis via FFT, and rigorous statistical testing with false discovery rate (FDR) correction.

Features

Automated batch processing of multiple trace files
Interactive neuron selection based on custom tags
Flexible time window selection for analysis
Polynomial detrending to remove slow drift
Autocorrelation analysis to detect rhythmic activity
FFT-based frequency analysis for oscillation characterization
Block-bootstrapping for robust statistical testing
FDR correction for multiple comparisons
Automatic figure generation and export

Requirements

MATLAB Version

MATLAB R2020b or later (uses argument validation syntax)

Required Toolboxes

Statistics and Machine Learning Toolbox (for mafdr)
Signal Processing Toolbox (for autocorrelation and FFT functions)

Installation

Clone this repository:

git clone https://github.com/yourusername/autocorrelation-pipeline.git

Add the repository to your MATLAB path:

addpath(genpath('path/to/autocorrelation-pipeline'))

Repository Structure

autocorrelation-calcium/
├── Main/
│   ├── Main_Autocorrelogram_script_standard_wrapped.m    # Main pipeline script for multifile analysis
│                              
├── Functions/
│   ├── data_extraction.m                                  # Load trace data and metadata
│   ├── select_neurons.m                                   # Interactive neuron selection
│   ├── time_selection.m                                   # Interactive time window selection
│   ├── polyfit_detrending.m                               # Polynomial detrending
│   ├── mp_autocorrelogram.m                               # Autocorrelation analysis
│   ├── autocorrelogram_freq_fft.m                         # FFT frequency analysis of the autocorrelogram
│   ├── shiftlinearcorr.m                                  # Block-bootstrapping statistics
│   └── blockbootstrp.m                                    # Block bootstrap resampling
└── README.md                                              # This file

Usage

Quick Start

Organize your data files with the naming convention: *traces.csv
Run the main script:

Main_Autocorrelogram_script_standard_wrapped.m

Select the folder containing your trace files
Follow the interactive prompts to select neurons and time windows

Input Data Format

The script expects CSV files with the following structure:

Time	Neuron_1	Neuron_2	Neuron_3	...
[Header Row]	tag_A	tag_B	tag_A	...
0.0	0.15	0.23	0.18	...
0.033	0.17	0.25	0.19	...
...	...	...	...	...

Row 1: Variable names (neuron identifiers) - becomes table header
Row 2: Neuron tags/categories for filtering
Row 3+: Time series data (first column = time, subsequent columns = neural traces)

Output

For each processed file, the pipeline generates:

Excel Results File (`Autocorrelation_results.xlsx`)

Contains the following columns:

Names - Neuron identifiers
max_autocorr_coeff - Maximum autocorrelation coefficient
oscillation_freq - Dominant oscillation frequency (Hz)
oscillation_freq_fft - FFT-derived oscillation frequency (Hz)
outside_CI - Bootstrap confidence interval test result
- 1 = above upper CI bound
- -1 = below lower CI bound
- 0 = within CI
lower_CI - Lower confidence interval boundary
upper_CI - Upper confidence interval boundary
p_value - Bootstrap p-value
BH_correction - Benjamini-Hochberg FDR significance (true/false)

Figures

All figures are automatically saved as PDFs in the same folder as input data:

traces.pdf - Raw and selected time windows
Detrending visualization
Autocorrelogram plots
FFT frequency analysis plots

Configuration

Main Parameters

Edit these variables in Main_Autocorrelogram_script.m:

% File selection
file_pattern = '**\*traces.csv';  % File search pattern

% Statistical parameters
fdr_correction = true;             % Enable/disable FDR correction
fdr_q = 0.05;                      % FDR significance threshold (default: 5%)

Advanced Parameters

Modify parameters in function calls for fine-tuning:

Bootstrap Parameters (line ~30)

[outside_CI(j), boundaries_CI(j,:), p_values(j)] = shiftlinearcorr(...
    detrended(:,j+1), detrended(:,1), coeff(j), ...
    "blocksize", 50, ...              % Block size for bootstrap
    "pValueCorrection", false);       % Conservative p-value correction

Additional parameters available in shiftlinearcorr.m:

CI - Confidence interval percentiles (default: [99 1])
iterations - Number of bootstrap iterations (default: 1000)

Detrending Parameters

Modify in polyfit_detrending.m function call

Autocorrelation Parameters

Modify in mp_autocorrelogram.m function call

FFT Parameters

Modify in autocorrelogram_freq_fft.m function call

Workflow

Input CSV Files
      ↓
[data_extraction]
      ↓
[select_neurons] - Interactive tag-based filtering
      ↓
[time_selection] - Interactive time window selection
      ↓
[polyfit_detrending] - Remove slow drift
      ↓
[mp_autocorrelogram] - Compute autocorrelation
      ↓
[autocorrelogram_freq_fft] - Frequency analysis
      ↓
[shiftlinearcorr] - Block-bootstrap testing
      ↓
[mafdr] - FDR correction
      ↓
Export Results & Figures

Statistical Methods